Beyond the Platform: Engineering Trust Across Digital and Physical Systems

Thank you, and good afternoon.

It's a privilege to close the Festival, and a slightly daunting one. Closing sessions are where half the room is checking flights and the other half is deciding whether it was all worth it. So I'll try to make the last thirty minutes count.

I should say at the outset that I'm speaking in my own capacity. I advise a number of companies in the technology and security space, but I'm not here for any of them.

I want to start with two words that are about as Australian as it gets. “No worries”.

It's our national shrug. Distance has been our comfort blanket: the lucky country, a long way from trouble. For decades, geography has shaped how we see the world. But the shifts of recent years mean we may need to make a bit of a fuss, even create a scene, to persuade those who make the decisions that we live in the world we have, not the one we might wish for.

Earlier this year, my colleagues at the ANU National Security College put a question mark after it. Their report, No worries?, draws on three national surveys of more than twenty thousand Australians, taken between late 2024 and early 2026, together with a series of focus groups.

The shrug has gone.

In November 2024, 42 per cent of Australians said they were worried about national security. By July last year, it was half. By February this year, it was 64 per cent. Almost two-thirds. Among 18- to 24-year-olds, worry went from 22 per cent to 55 per cent in fifteen months.

So what worries them the most? AI-enabled attacks. A severe economic crisis. Critical supply disruption. Disinformation. Foreign interference. And of the fifteen threats Australians were asked to rate, a foreign military attack came last. By February this year, concern about cyber threats had reached 78 per cent.

These concerns are real, and they're lived. The scam that empties a pensioner's account. Stolen personal data. Ransomware. Deepfakes. The fraud that starts in a chat app and ends in a cryptocurrency transfer. Coordinated campaigns by state actors. The list goes on. Australians are telling us that the threats they fear most move through the very systems many of you have spent your careers trying to make safe.

So, a small confession. I'm a practitioner at heart: policing, national security and the border. Which probably makes me the least likely speaker you've heard over these two days. I don't moderate content. I don’t design digital architecture, and I've never written a platform policy. But I have spent 44 years in the trust and safety business. Just on different kinds of platform.

I spent over seven years leading the Australian Border Force. I was responsible for a system that handles tens of millions of travellers and visa holders, and more than a billion cargo transactions a year, and that watches over a maritime domain covering more than eight million square kilometres. Most of it runs on trust: in data, in declarations and in risk scores, because physically checking everything is impossible. We had false positives that hurt legitimate travellers and traders, and false negatives that hurt everyone. We had adversaries who adapted faster than our rules could, and a public that, quite reasonably, expected us to get it right.

We also lived with an old bargain: more security meant more friction, and faster trade meant more risk. In your world, I suspect it's safety versus growth. So the question I want to put to you today is whether that bargain still holds, and in the world we now live in, whether it's even viable.

Over the next half hour, I want to look at the border as a system, rather than a line. I'll walk through one real case, illicit tobacco, that shows how harm moves through the gaps between physical and digital systems. I'll talk about how we engineer trust across those gaps and be honest about the risks of doing so. And I'll finish by coming back to that old bargain, and why I think there's real cause for hope.

The border is a system of physical and digital interfaces

Ask most people what the border is and they'll describe the parts they've encountered. A SmartGate. An incoming passenger card. A detector dog at the baggage carousel.

These touch points are the smallest components of a complex system that underpins our economy, our productivity and our security.

The border is like a chain. For trade, it starts where goods are made, ordered, financed and documented, often in another country, long before anything or anyone moves. It runs through ships, planes, ports, freight forwarders, warehouses, banks and payment providers. And it continues downstream, into shops, homes and bank accounts, long after anything has cleared.

Every link in that chain trusts the link before it. The port trusts the shipping line's manifest. The shipping line trusts the forwarder. The forwarder trusts the importer. And the regulator trusts all of them, most of the time at least, because it has to.

That's the model most platforms run on too. Trust by default. Verification by exception.

None of that will surprise anyone in this room. You've lived in a world where the physical and the digital merged a long time ago. What may be less obvious is where those chains now begin. The ad, the order and the payment increasingly happen online, before a container is ever loaded. Which means many of the chains that end at our ports now start on digital platforms. Whether we planned it or not, those platforms, which many of you help keep safe, have become the first link in the border.

That’s important, because our institutions weren't built for it. They were built to say “yes” quickly: to clear, process and move things on. The platform takes the order. The bank processes the payment. Customs clears the container. Police respond to the firebombing. Each is organised around its own mandate, measured on its own KPIs and accountable for its own slice. Everyone owns a transaction. Who owns the harm? Too often, nobody.

And the machinery that should connect them hasn't kept pace. Governance is fragmented, with no one responsible for the system as a whole. The technology is ageing: our main cargo system is more than twenty years old, and it was never designed to share data securely, let alone enable machine learning. And the networks that do connect the various players in the system depend too often on who knows whom. That can work okay in a crisis, mostly. But it doesn't scale.

Twenty years ago, the Auditor-General found that Customs could examine around seven per cent of sea containers. Trade volumes have grown dramatically since then. Today, it's one or two per cent. That isn't a criticism of the officers. They do extraordinary work. It's the arithmetic of volume. And those volumes are projected to keep growing.

But the answer isn't to accept the status quo. We can’t hire enough officers to open every container, any more than you could moderate a platform by having a human read every post. What we can do is combine better signals with better tools, and place human judgement where it counts most. The border is the same. The technology now exists to scan cargo as it moves through the supply chain, and for AI to read that image alongside everything else we know: the declaration and bill of lading, the shipping and sensor data, and information held across industry and government. That's what tells an officer where to look. See more. Screen more. And leave the decisions that matter with people.

Visibility and precision are the decisive capabilities. Let me show you what happens without them.

A real-life study: illicit tobacco

Let me make this concrete with a case that I suspect has touched every community represented here.

Illicit tobacco.

In 2017, around one in every eight cigarettes smoked in Australia came from the black market. By last year, according to the ABS, it was four in every five. Australia's Illicit Tobacco and E-Cigarette Commissioner estimates the combined illicit tobacco and vape market at between 5.7 and 8.5 billion dollars a year, with foregone excise of up to 11.8 billion dollars. Tobacco excise collections have all but collapsed, from a peak of 16.3 billion dollars in 2019–20 to just 3.9 billion dollars last financial year. It's now one of the most lucrative criminal markets in the country.

Follow a packet through the system and you'll see why it's so hard to stop.

It's advertised on social media and ordered through an encrypted messaging channel. It's often paid for in cryptocurrency or through a payment app. It arrives in a container, under a manifest that describes something else entirely. It's cleared, because it's one of the vast majority of containers that are never physically examined. It's broken down in a warehouse, distributed to a shopfront, and sold over the counter. And the proceeds, billions of dollars a year, are laundered: converted into cryptocurrency, or dressed up as payment for a shipment of goods that never existed.

Almost every step of that journey leaves a digital trace: the ad, the account, the payment, the manifest, the shipping movement and the cash deposit.

Most of the data is there. What's missing is the architecture to connect it, lawfully and at speed.

What the border misses, the community pays for. Harm flows downstream: into the legitimate businesses it undercuts, the health system that picks up the cost, and the communities that live with the violence around it. In Melbourne, well over a hundred premises have been firebombed in the fight for control of the illicit market, and lives have been lost, including a young woman who died in a house fire in what has been reported as a case of mistaken identity.

Remember what the No worries? research found Australians value most. It isn't "national security" in the abstract. When asked to choose the nation's top priority, the clear winner was safe and peaceful communities. Nearly two-thirds chose it as their first or second preference, and that finding held largely across age, income and location.

Firebombed shops on suburban shopping strips are the precise opposite of that.

And it goes further. In 2023, at the request of the United States Department of Justice, Australian authorities arrested a man living in Sydney who, according to US prosecutors, ran a tobacco smuggling network that generated hundreds of millions of dollars for North Korea: money that prosecutors allege helped fund the regime's weapons programs. The cigarettes moved through the same trade corridors as everything else. The same containers. The same declarations. The same ports.

This is where our neat lines of responsibility break down. A cigarette smuggling network that helps fund a nuclear-armed regime is organised crime, sanctions evasion and a national security threat. It's also a tax problem, a health problem and a platform problem. And because it's all of them, no single agency, and no single company, owns it.

So this isn't simply an intelligence failure. Everyone knows the illicit tobacco is there: it's sold openly on shopping strips across the country. Much of the information is there too. What's missing is the structure to act on it. Harm flows most easily through the seams: between systems, between agencies, between jurisdictions, between the physical and digital worlds. Because that's where nobody sees the whole picture, and nobody owns the harm.

Most of you will recognise the pattern elsewhere. The scammer who opens the conversation on one platform, moves it to another, and takes the money through a third (Increasingly, the person typing those messages has been trafficked and is being held in a scam compound offshore). The offender banned here who reappears over there. The extortion crew that steals data in one country, leaks it on a forum hosted in another, and demands payment in crypto. Illicit tobacco is simply the physical manifestation of a problem many of you deal with every day.

Extending trust across the gaps

So what do we do about it?

When things go wrong or get hard, government's instinct is often to centralise. Create a super-agency, pool all the data and wrap it in a giant system.

That instinct is usually wrong, for practical, legal and democratic reasons. A single central database of everything is expensive, brittle and a magnet for attack. And it would struggle to earn, or deserve, public trust.

The better model is a federated architecture. Each organisation keeps control of its own data, but the systems are linked so that risk signals can be shared lawfully, in near real time, across the chain. The algorithm travels to the data. The insight is shared, while the raw data, and people's privacy, stay protected.

That can sound abstract, so let me give you two examples of it working.

The first is from medicine. Hospitals have been training AI models to read CT scans, MRIs and X-rays across multiple institutions for some time, without ever sharing a single patient record. Each hospital trains the model locally, and only the learning, the model updates, is exchanged. Patient data never leaves the building. If that level of trust is achievable for cancer diagnosis, under some of the strictest privacy regimes in the world, it is surely achievable for trade.

The second example is closer to home for most of you.

I won't try to explain these to this room, because most of you will know them far better than I do. But from where I sit, GIFCT's hash-sharing database, StopNCII and the Tech Coalition's Lantern program all work on the principle I've just described: the risk signal is shared, while the sensitive material and the decision stay where they belong. And this year, Lantern was opened to financial institutions worldwide, bridging the seam between platforms and payments. That's the same kind of gap illicit tobacco exploits: between where deals are done and where money moves. These programs show the model can work, but the border hasn't caught up.

So what would it look like at the border? My National Security College paper earlier this year, Beyond the Checkpoint, argued for a National Border Visibility Platform: not a super-database, but a federated backbone, built on common standards and secure interfaces, that lets government agencies and trusted industry partners share risk signals in near real time, while data stays with its lawful owner. I see it as a key part of our national resilience, not just our border security.

We've already had a glimpse of what that can do. In late 2022, the Australian Border Force co-organised Operation Tin Can with the World Customs Organization and the UN Office on Drugs and Crime. Fifty-eight countries, supported by major shipping lines, lawfully combined industry data with law enforcement intelligence. In under four weeks, they seized around 98 tonnes of cocaine hidden in legitimate supply chains, much of it concealed with the help of corrupt insiders. It was a single operation, but it proved the concept for a permanent capability that has yet to be built.

A federated architecture opens up other possibilities too, including the ability to test before we deploy. Singapore is building a digital twin of its port and maritime system: a live computational model of how ships, goods and data actually move. Pair a twin like that with a digital twin of our policies, the rules themselves expressed as code, and you can ask the question anyone changing a policy should ask: what happens when we switch this on, to legitimate traffic, in a crisis, and to where the bad actors go next? At the border today, we mostly find out after the fact, if we find out at all.

And there's an investment argument. When a port operator or logistics company invests in better supply-chain data, it creates value for everyone, through safer trade and sharper targeting, but it captures almost none of that value itself. That's a market failure, and it leaves private capital on the table that could be doing public work. If those who invest in visibility earned faster clearance and fewer holds, what I call assured trade, the business case for safety would write itself. Some of you will recognise that argument: safety as a competitive advantage, rather than a cost of doing business. Today, only a small fraction of Australia's traders are accredited as trusted traders. Earlier I described a system built on trust by default. The better model is trust earned through visibility, with attention going to wherever visibility is poor.

Being honest about the risks

Now, I'd be doing you a disservice if I stopped there. Everything I've just described creates risks of its own, and this audience, you of all audiences, I’m sure will want me to describe them.

First, cyber. Connect systems and you create new pathways for attack. A national visibility platform would be critical infrastructure in its own right, a target for criminals and hostile states, and it would have to be designed and protected as such from day one.

Second, the concentration of insight. Even if raw data never moves, the insights do, and a map of how risk is scored is extremely valuable to the people trying to evade it.

Third, the adversary has AI too. The No worries? research found that just 4 per cent of Australians think the nation is very or fully prepared for AI-enabled attacks. Criminal networks and hostile states are already using AI to generate synthetic identities and to probe our defences. Anything we build has to assume it will be tested by adversaries using the same tools.

Fourth, insiders. After the covert Operation Ironside, our follow-up work at the border identified around a hundred companies and a thousand individuals working in our ports, airports and supply chains with links to organised crime. Systems that assume good faith can be defeated from the inside. Trust has to be continuously verified, not granted once.

And fifth, the risk you know best: mission creep and the erosion of privacy. Tools built to stop smugglers must never become tools for watching everyone. And false positives can harm real people.

So the safeguards have to be engineered in from the start, not bolted on after the first catastrophe. That means data stays with its owner, purposes are defined and limited, every query is logged and auditable, consequential decisions stay with people, and there is independent oversight with real teeth.

We already know what the standard should look like. In healthcare, AI used in diagnosis increasingly has to be regulatory grade: accurate, explainable, tested in real-world conditions and independently assessed before it's deployed. AI at the border should meet the same bar. And agencies should be open about how it performs: how often it's right, how often it's wrong, and who it gets wrong. Many of you already publish transparency reports on those kinds of questions. A national border platform should be held to no lower a standard.

Trust, and the social licence that comes with it, has to be engineered as well as earned.

The social licence

That brings me back to the Australian public.

When I wrote Beyond the Checkpoint, I made a recommendation I felt strongly about. I noted that there appeared to be no national research into what Australians would accept when it comes to border modernisation: shared data, digital infrastructure, AI-supported decisions. I argued that, without understanding community expectations and trust thresholds, even the strongest operational case risks being politicised.

Two months later, No worries? was published. It doesn't answer my question directly. But it tells us a great deal about the ground we'd be building on.

In the focus groups, security agencies were generally described as credible and professional. More than half of Australians think government tells them too little about the threats we face. Only 4 per cent think they're told too much. And on every one of fifteen threats, fewer than one in five believe the nation is well prepared.

That's a public that wants more capability, not less, and wants to be in on the giggle.

And there's one more finding that I think matters most for this room. The researchers found that many Australians want to contribute to national security but don't know how. Surveyed after the Bondi attack in December, 71 per cent agreed that all Australians can do more to make our communities peaceful and safe. Only 8 per cent disagreed.

People want to help, but don't know how. That's a design problem. And trust and safety professionals are, at heart, designers of the channels through which ordinary people can report, flag, verify and protect. You build the bridges between individual vigilance and system response. At the border, that kind of design still has a long way to go.

Engineered hope

Jenine asked me to leave you with the message that it's not all gloom and doom. I'll confess that request gave me pause. In March, opening a conference of policy innovators in Canberra, I said, and I believed it, that hope is neither a plan nor a strategy. I still believe that. But I also know how easy it is to become cynical in policing and national security, and how important it is that we don't: that we bring solutions, and bring them with optimism. So let me reconcile the two.

The hope I'm offering isn't wishful. It's engineered. It rests on things that already exist, and already work.

The technology exists. Federated learning is protecting patient privacy in hospitals today. Hash-sharing is keeping terrorist content and abusive images off platforms today. Digital twins are already being used in major ports today.

The proof of concept exists. Fifty-eight countries, working with a group of shipping lines, seized 98 tonnes of cocaine in under a month, simply by sharing new data and illuminating what they couldn’t see previously.

The partnership model exists. AUSTRAC’s Fintel Alliance has had banks, regulators and law enforcement working together on financial crime for almost a decade, whilst protecting sensitive information and without everyone pooling their data.

And the old bargain no longer holds. Earlier I asked whether it still does. My answer is no. With better visibility, throughput and assurance can rise together, and the investment case and the security case point the same way. In public policy that's rare, and we shouldn't waste it.

And the public is ready. Not complacent. Concerned, wanting to know more, and wanting to help.

What's missing isn't capability. It's connections: across agencies, across sectors, and across the lines between the physical and the digital. And that's where this room comes in. Trust and safety is a young profession, but it's already the one that lives in the gaps, between technology, policy, law and the front line. Many of the people who will engineer trust across those seams are sitting in this room.

So let me come back to where I started.

No worries?

The honest answer is no. Australians do have worries, and they have good reason for them.

But worry isn't the opposite of hope. Complacency is. And the research tells us this country is not complacent.

So my answer to that question mark isn't "no worries". It's this: worries, and the tools, the people and the will to act on them.

We built these systems, and we can rebuild them. Trust hasn't been lost. It's under construction, and many of you are already building it.

Thank you.